← All workOpen source · 2026

LinkedIn Puzzle Solvers

A browser extension that reads LinkedIn's daily puzzle boards with computer vision, solves them and applies the answer. Six games, local or hosted.

6
daily games solved
2
browsers
Fig. 1 · The v0.3.0 popup, and the live API's solution previewed on a Queens test board

LinkedIn publishes a handful of small logic puzzles every day: Queens, Tango, Mini Sudoku, Zip, Patches and Wend. This project solves all six from the page itself.

The browser extension (Chrome and Firefox) captures the board and sends the image to a FastAPI solver service. Each game has its own parser that turns pixels into a grid with OpenCV and OCR, and its own solver. The extension then applies the solution back on the page.

The six daily games, all parsed and solved from the page by the extension

How it’s built

  • Two ways to run it: Online, against the hosted API on Vercel, or Local, with the same API in Docker on 127.0.0.1.
  • Hexagonal boundaries between the API, the per-game parsers and the solvers, so a new game is a new adapter, not a rewrite.
  • CI with regression tests on saved boards, Gitflow and versioned releases shared by the extension, the API and the Python package.
Case studyRepoNoteTF-IDF + PCA · 25 documents

Search by project name or topic↑↓ move ↵ open